Quarterly landings or catches of total fishes and the major pelagic fish species, were forecasted using the methods and models viz. autoregressive integrated moving average (ARIMA), non-linear autoregressive (NAR) artificial neural network (ANN), autoregressive integrated moving average with exogenous inputs (ARIMAX), non-linear autoregressive with external (exogenous) inputs (NARX) artificial neural network. The models were also developed by considering only two important variables (differ for total fish and selected fish species) obtained from the ANN model. These simplified models proved nearly as good in their predictions. Simulated sea surface temperature (SST) for the A2 climate change scenario was used as an input for the NARX model to estimate the catches of Bombay duck over a short term (2020 – 2025) and a long term (2030 – 2050) with the last two years’ (2012 – 2013) average catch of training data as a benchmark. The catches increased on average by 41 % in the short term but decreased by 17.72 % in the long term.

中文翻译：

---

预测总鱼类和主要中上层鱼类的季度上岸量，并模拟气候变化对印度西北部古吉拉特邦海岸孟买鸭的影响

使用方法和模型，即预测总鱼类和主要中上层鱼类的季度上岸量或捕获量。自回归综合移动平均 (ARIMA)、非线性自回归 (NAR) 人工神经网络 (ANN)、具有外源输入的自回归综合移动平均 (ARIMAX)、具有外部（外源）输入的非线性自回归 (NARX) 人工神经网络。这些模型也是通过仅考虑从 ANN 模型获得的两个重要变量（总鱼和选定的鱼种不同）而开发的。这些简化的模型在他们的预测中被证明几乎一样好。A2 气候变化情景的模拟海面温度 (SST) 被用作 NARX 模型的输入，以估计短期（2020 年至 2025 年）和长期（2030 年至 2050 年）孟买鸭的捕获量，最后两年（2012 年至 2013 年）的平均训练数据捕获量作为基准。捕捞量短期内平均增加了 41%，但长期减少了 17.72%。